Chips · Oct 21, 2026, 18:43
Dell Packs 256 NVIDIA Vera CPUs Into One 48U Rack, Due Q1 2027
Dell's PowerEdge M9882 rack holds 256 Vera CPUs for agent workloads, while the XE9810 adds 16 liquid-cooled Groq 3 LPU trays; both ship in Q1 2027, with no power figures yet.
A Dell integrated rack from the Dell AI Factory with NVIDIA page; it is not the PowerEdge M9882 or XE9810 themselves.
Image: Dell Technologies
Source · Press image from the Dell Technologies website, used for news reporting.
Dell Technologies said on October 21 that its PowerEdge M9882 server packs 256 NVIDIA Vera CPUs into a single 48U rack, and that the PowerEdge XE9810 will be the first to deliver rack-scale inference built from 16 direct-liquid-cooled compute trays powered by Groq 3 Language Processing Units (LPUs). Dell announced both in a Business Wire release from NVIDIA's GTC event in Berlin, and both are due in the first calendar quarter of 2027.
Dell’s two new rack systems, the 256-CPU M9882 and the Groq 3 XE9810, arrive in Q1 2027, months after the desktop and R8822 parts.
Illustration (AI-generated)
The M9882 is aimed at the CPU side of agentic AI: orchestration, tool use, code execution and evaluation. Dell said that splitting this work from GPU clusters lets operators avoid over-provisioning mixed-use servers and spend their data center power and cooling capacity where it counts. The XE9810 pairs with existing GPU clusters as what Dell described as a disaggregated complement for high-throughput inference.
Sources
- Dell Technologies Advances Enterprise Agentic AI from the Deskside to Data Center with NVIDIA · Google News · chip power